JAPANESE INPUT METHOD
INDEPENDENT OF APPLICATIONS
By Takahide Honma, Hiroyoshi Baba, and Kuniaki Takizawa
ABSTRACT
The Japanese input method is a complex procedure involving
preediting operations. An application that accepts Japanese from
an input device must have three systems for the input method: a
keybinding system, a manipulator for preediting, and a
kana-to-kanji conversion system. Various keybinding systems and
manipulators accelerate input operations. Our implementation
separates an application from the Japanese input method in three
layers. An application can use a front-end input processor to
perform all operations including I/O. An application can use the
henkan (conversion) module and implement I/O operation itself. An
application can execute all operations except keybinding, which
is handled by an input method library.
INTRODUCTION
In this paper, we first present an overview of the technical
environment of the Japanese input method implementation. Based on
this overview, we briefly describe the critical engineering
issues for conversion of Digital's products for the Japanese
user. Our most critical engineering issue was the reduction of
similar (but slightly different) work to localize products.
Another issue was to satisfy customers' requests for the ability
to use the many input styles familiar to them. We describe our
approach to the development of a Japanese input method that
overcomes these issues by separating the input method from an
application in three layers.
OVERVIEW OF THE JAPANESE INPUT METHOD
In this section, we describe Japanese input and string
manipulation from the perspective of both the user and
application. Based on these descriptions, we present a
overview of reengineering a product for Japanese users
summary of the industry's complex techniques developed
Japanese input methods.
the
brief
and a
for
Japanese Input
The Japanese writing system uses hundreds of ideograms called
kanji. In addition, Japanese uses a phonetic system of
kana characters (hiragana and katakana) and has accepted
romaji, which is the use of Latin letters to spell Japanese
words. Figure 1 summarizes the Japanese character systems.
Japanese input requires users to operate in a "preediting" mode
to convert kana or romaji into a kana-kanji string.[1,2]
Figure 1 Japanese Character Systems
JAPANESE CHARACTERS ----|
|
|
|
|
|
|
|
|
|
+--
PHONOGRAM
|
+-- KANA (ORIGINAL JAPANESE CHARACTERS)
|
|
|
+-- HIRAGANA
|
|
|
+-- KATAKANA
|
+-- ROMAJI (USING LATIN ALPHABET TO EXPRESS A
PHONEME)
IDEOGRAM
|
+-- KANJI
The computer keyboard used for Japanese input has multiple
character assignments. Almost all keys on the Japanese keyboard
are assigned both a Latin alphabet character and a Japanese kana
character. The Japanese user must first choose between kana key
input or alphabet input. A user in an engineering area generally
uses romaji (alphabet) key input. In the office environment,
however, a user prefers kana key input because it requires half
as many keystrokes as romaji input.
Preediting Operation
The user inputs the phonetic expression in either kana or romaji
that represents the statement the user wants to input. Then the
user presses the conversion key to convert the phonetic
expression to a kana-kanji mixed string. At this time, the user
checks the accuracy of the conversion result. Sometimes the user
needs to select the correct word from a system-generated list of
several homonyms. Moreover, a user may also need to determine the
separation positions in the phonetic expression to ensure a
meaningful grammatical construction.
Japanese has no word separator equivalent to the space character
in English. To obtain the correct or expected separation of
grammatical elements, the user must sometimes move the separation
position. After the user constructs a corrected statement, he or
she finishes preediting and fixes the statement. The user repeats
these complex steps to construct Japanese documents. Figure 2
shows the preediting steps for the Japanese user.
Figure 2 Preediting Japanese Input
START
| SET UP INPUT METHOD
|
----> INPUT PHONETIC EXPRESSION FOR A STATEMENT
| |
|
| |
+-- (CHANGE PHONOGRAM SYSTEM)
| |
|
| |
+-- CONVERT KANA-TO-KANJI
| |
|
|
| |
|
+-- MOVE THE SEPARATION POSITION
| |
|
|
| |
|
+-- SELECT WORD FROM MANY HOMONYMS
| |
|
| |
+-- FIX A STATEMENT
| |
|
| +----+
+-- END OF DOCUMENT
Various techniques have been developed to accelerate Japanese
input operations. They include UNDO, COPY, zip code conversion,
and categorized expert dictionary.
Japanese Application Capabilities
The Japanese application has two special capabilities for
Japanese processing. First, the application must be capable of
handling multibyte characters. This subject itself is interesting
as it involves wchar_t and Unicode character sets; however, this
paper focuses on the second capability, the implementation of the
input method. An application that accepts Japanese from an input
device must have, at least, three additional systems for the
input method. These are the so-called keybinding system, a
manipulator for preediting, and the kana-to-kanji conversion
system.
Keybinding System. This system analyzes the key input from a
user and determines which of the key's functions the user wants
to do. It defines the user interface and the way a user operates
with keystrokes. It also defines the preediting conversion key.
We imagine there are as many keybinding systems as there are word
processors.
Preedit Manipulator System. This system not only echoes the
input characters on the screen but also controls the video
attribute that expresses the preediting area. This capability
allows the user to distinguish preediting strings from background
fixed strings. The user must be able to recognize the preediting
string for more processing (for example, to convert the input to
another expression such as kana to kanji). In addition, the user
can set this system to convert input to another expression
dynamically (for example, automatic conversion of romaji to
kana).
Kana-to-kanji Conversion System. This system analyzes the input
string, gets the word from a dictionary, and constructs the
correct statement grammatically. Many personal computer (PC)
vendors have invested in systems that use this input method. In
Japan, some vendors have introduced artificial intelligence
technology, but this system essentially has only statistical
rules.[3,4]
Figure 3 summarizes Japanese processing as handled by
applications.
Figure 3 Japanese Processing by Applications
JAPANESE PROCESSING ----- VARIABLE MULTIBYTE CHARACTER HANDLING
|
+-- JAPANESE INPUT METHOD
|
+-- KEYBINDING SYSTEM
|
+-- PREEDIT MANIPULATOR SYSTEM
|
+-- KANA-TO-KANJI CONVERSION SYSTEM
Method of Japanese Conversion
As mentioned above, to convert a product for use in Japanese, we
must implement both a Japanese-string manipulator and an input
method. To retain the "look-and-feel" of the original Digital
product, the interface is designed so the Japanese user does not
need to explicitly enter the preediting session with the
special-purpose key (Enter Preedit) but is automatically entered.
With most applications on other systems, a user must explicitly
enter the preediting session by using the special keystroke. This
implementation has the advantage that it completely separates the
input method from the application, but it requires the user to
remember to perform an extra step.
To eliminate the conflict between the original product's key
function and the additional Japanese input function, each product
has to have a slightly different keybinding system for Japanese.
As a result, a user must learn more than one Japanese input
operation when using multiple products.
User Environments
PCs are widely used in many offices and are popular devices for
Japanese input. Naturally a user wants to operate with a familiar
PC keystroke for Japanese input even in integrated systems (in
some servers). When PCs, which use front-end processors, are
integrated into environments with VMS and UNIX systems, a user
often prefers the PC's interface. The more integrated a user's
environment is, the more requirements a user has.
In addition, a distribution kit for the X Window System in a UNIX
environment has some sample implementations of the Japanese input
method. This kit gives a user more choices for input at no cost.
The market for the Japanese input method separates vendors into
two main groups. One is the PC front-end processor manufacturer
who implements more advanced techniques but at a high price. The
other is the UNIX system vendor who supplies input
implementations free (without guarantee) and thus reduces the
maintenance cost.
In the next section, we present our approach to the development
of an application-independent Japanese input method. The goals of
our design were (1) to include the PC keybinding system in
integrated environments so users could select their preferred
input method, (2) to supply a tool that would easily convert
products for the Japanese user, and (3) to provide a way to
access the interfaces of several Japanese engines and thus
capture the free software capabilities.
APPLICATION-INDEPENDENT APPROACH
As described in the previous section, the Japanese input method
includes complex techniques. Many PC software vendors (but not
manufacturers) decided against developing their own methods and
incorporated a popular input method for their applications. This
decision, of course, reduces their development cost. Our approach
also seeks to reduce development cost. We separated the input
method from the application to the greatest extent possible, as
long as the separation did not adversely affect the application.
The PC system is designed as a single-task system, but Digital's
operating systems (OpenVMS, ULTRIX, and DEC OSF/1 AXP) are
designed as multitasking systems. Therefore we could not adopt
many of the PC techniques that were implemented in the driver
level. For example, access to dictionary and grammatical analysis
of the input string are too expensive in the driver level of a
multitasking system because they use system resources that are
common to all tasks on the system.
Our approach divides the input method into three layers. Each
layer is dependent on any lower layer. Consequently, any
application using the highest layer also uses the functionalities
of the other two layers.
Strategy of Three Layers
The criteria of our layering strategy were (1) to reduce the cost
of reengineering products for the Japanese user, (2) to unify the
input method user interface, and (3) to protect the user's
operational investment in a keybinding system.
These criteria led us to set the keybinding system into the
lowest layer. We named our system the input method library
(IMLIB) and released it on VMS/Japanese version 5.5 and
ULTRIX/Japanese version 4.3. We also ported it to the Alpha AXP
system, and this facility is available on any Japanese platform.
Any application using our method needs to use IMLIB.
In essence, this keybinding system allows a user to change the
input method of operation to any style by changing the IMLIB
definition files. If an application supports IMLIB, a user can
change the application's input operation by changing IMLIB once.
As a result, an application's key customization function can move
into IMLIB.
At this point, we considered the simplest method of separating
the input method from applications. The intermediate process,
also called the front-end method, processes all the input and
then passes it to an application. Many front-end implementations
use the pseudo-terminal driver (pty in UNIX or FT in OpenVMS).
The intermediate process gets all I/O to and from an application,
processes it, and finally passes it to an application or a
device. This implementation cannot recognize the application
input requests, but works only by a user's operation. To change
this operation, we set the hook inside the terminal driver to get
all application-request information. Our front-end process
recognizes application requests and works without conflict.
One advantage of this front-end implementation is a complete
independence of applications. This can also be a disadvantage
since an application cannot control the input method closely. For
example, this implementation can alter the user interface of an
editor system.
We continued to study another layer for separation. The
preediting operation, that is, all the input string manipulations
except I/O to devices, was a candidate. All applications pass the
input from input devices into the Japanese input manipulator and
then pass the output from this manipulator onto output devices.
By using this system, we can unify the input operation except for
device-dependent operations and reduce the cost to implement this
kind of functionality.
Our development process started at the lowest layer (IMLIB),
proceeded to the highest layer (front-end), and finished at the
middle layer (preediting manipulator). In the following sections,
we describe the functionalities in each layer from the lowest to
the highest layer.
IMPLEMENTATION OF IMLIB
IMLIB is a utility that supplies the keybinding definition
function and other information for customizing the Japanese input
operation. This capability enables us to supply user-friendly
keybinding systems. A user can change the input sequences and the
look-and-feel of the user interface by modifying databases. We
introduced two databases, KEYBIND for keybinding and PROFILE for
look-and-feel and an application's usage. We also supplied the
KEYBIND compiler for improved performance and the elimination of
the grammatical error at run time.
As mentioned in the introduction, there are many implementations
of Japanese input styles on PCs or some word processors, and some
text editors on various operating systems. If a user needs to use
a different editor, he or she needs to learn another operation.
Our method unifies the input operation. We studied several types
of input styles and recognized that we could build the general
model for this input operation. The IMLIB manual describes this
model in detail.[5,6] In this paper, we discuss it briefly.
KEYBIND Database
In the Japanese input operation, entering the key input causes
several conversion actions and state transitions. Figure 4 shows
the multiple transitions incurred during input. We needed to
define the conversion actions and some state transitions as a
single key input action. We implemented this function through the
KEYBIND database and language. Figure 5 is an example of the
KEYBIND database. A user builds an input style by changing the
KEYBIND database with the KEYBIND language.
Figure 4 State Transition
+--------------+
+---------------+
|INITIAL STATE |<--------->|KANJI CONVERTED|
+--------------+ <
>+---------------+
^
\
/
^
|
\ /
|
|
\/
|
|
/\
|
|
/ \
|
v
/
\
v
+---------------+ <
>+---------------+
|INPUTTING STATE|<------->| KANA CONVERTED|
+---------------+
+---------------+
Figure 5 Portion of KEYBIND Database
! (JVMS conversion key definition file (system template) ver 1.0)
!
gold
= CTRL_G;
! Gold key; used as a PREFIX key
kakutei
= CTRL_N;
! Finish without any conversion
kanji_henkan
= NULL, gold + CTRL_K; ! Convert to Kanji / next
candidate
hiragana_henkan = CTRL_L;
! (Convert clause) to Hiragana
katakana_henkan = CTRL_K;
! (Convert clause) to Katakana
zenkaku_henkan = CTRL_F;
! Convert to full width characters
hankaku_henkan = gold + CTRL_F;
! Convert to half width characters
kigou_henkan
= GS;
! Symbolic code conversion
oomoji
= VOID;
! Convert to upper characters
komoji
= VOID;
! Convert to lower characters
ji_bunsetsu
= CTRL_P;
! Move to next clause
zen_bunsetsu
= gold + CTRL_P;
! Move to previous clause
tansyuku
= US;
! Shrink the clause
sintyou
= gold + US;
! Extend the clause
zen_kouho
= gold + (NULL, CTRL_L);! Previous clause candidate
kaijo
= CTRL_N;
! Cancel the conversion
!
and go into input state
sakujo
= DEL;
! Delete previous character
hidari
= LEFT;
! Move the cursor left
migi
= RIGHT;
! Move the cursor right
space_first
= "\ ";
! Finish by space
! (space at initial state)
space_input
= "\ ";
! Finish by space
! (space at other states)
STATE "initial" =
space_first
kanji_henkan
hiragana_henkan
katakana_henkan
zenkaku_henkan
hankaku_henkan
kigou_henkan
oomoji
komoji
TYPING_KEYS
END;
...
:
:
:
:
:
:
:
:
:
:
NONE;
START_SELECTED, CONVERT, GOTO "kk_converting";
START_SELECTED, HIRAGANA, GOTO "converting";
START_SELECTED, KATAKANA, GOTO "converting";
START_SELECTED, ZENKAKU, GOTO "converting";
START_SELECTED, HANKAKU, GOTO "converting";
START_SELECTED, SYMBOL, GOTO "converting";
START_SELECTED, UPPER, GOTO "converting";
START_SELECTED, LOWER, GOTO "converting";
START, ECHO, GOTO "inputting";
IMLIB allows the user to change the keybinding and to choose a
different input sequence with a different state transition
vector. For the user's convenience, IMLIB provides some KEYBIND
databases of the major Japanese input styles in default.
When an application calls the ImSetKeybind function, it loads a
KEYBIND binary file into memory. Each time the application gets
the key input, it queries the key's function from IMLIB. IMLIB
searches the KEYBIND file for the key's definition and returns
that information, called an action, to the application. Each
action is a set of orders that has a different procedure for
Japanese conversion. For example, the action CONVERT means to
convert an input string to a kanji string. At that time, IMLIB
also maintains Japanese input states and, if necessary, changes
the state.
PROFILE Database
The Japanese input operation has many parameters to determine its
look-and-feel, such as the video attribute for the preediting
string, preediting exception handling, and application-specific
processing. The PROFILE database stores these additional
parameters the same way as the resource file does in the X Window
System.
The PROFILE database is a text file. It contains several records
that represent each environment. This record format has the style
of INDEX : value. The application predefines the INDEX for its
purpose; however, IMLIB defines some INDEXes related to Japanese
input operation because it requires some common environment
definitions. The range or value corresponding to the INDEX is
placed in the right-hand side of the record. Figure 6 shows a
record from a PROFILE database.
Figure 6 PROFILE Database Record
DEC-JAPANESE.KEY.keybind : im_key_jvms_level2
DEC-JAPANESE.KEY.keybind_1 : im_key_jvms
DEC-JAPANESE.DISP.preEditRow : current
DEC-JAPANESE.DISP.preEditColumn : current
DEC-JAPANESE.DISP.inputRendition : bold
DEC-JAPANESE.DISP.kanaRendition : bold
DEC-JAPANESE.DISP.currentClauseRendition : reverse
DEC-JAPANESE.DISP.leadingClauseRendition : none
DEC-JAPANESE.DISP.trailingClauseRendition : none
DEC-JAPANESE.ECHO.ascii : hankaku
DEC-JAPANESE.ECHO.kana : hiragana
DEC-JAPANESE.ECHO.autoRomanKana : off
DEC-JAPANESE.OUTRANGE.clauseSize : none
DEC-JAPANESE.OUTRANGE.clauseNumber : rotate
DEC-JAPANESE.OUTRANGE.cursorPosition : done
KEYBIND Compiler
The KEYBIND compiler analyzes the KEYBIND text file and creates
the KEYBIND binary file. IMLIB services reads the PROFILE
database and the KEYBIND binary file and maintains them in
memory. As a response to an application's query, IMLIB services
sends it the actions in KEYBIND and the data in PROFILE and at
this time maintains the KEYBIND states. Figure 7 shows the
relationship among the IMLIB components.
Figure 7 Relationship among IMLIB Components
+-------------+
| APPLICATION |
+-------------+
|
^ ACTIONS (KEYBIND)
QUERY v
| KEYWORDS (PROFILE)
+----------+
+----------+
| KEYBIND |
| IMLIB
|
| COMPILER +--------+ +-------->| SERVICES |
+----------+
| |
+----------+
^
| |
^
|
v |
|
/~~~~~~~\
/~~~~~~~\
/~~~~~~~~\
|KEYBIND|
|KEYBIND|
|PROFILE |
|TEXT
|
|BINARY |
|DATABASE|
|FILE
|
|FILE
|
|FILE
|
\_______/
\_______/
\________/
IMLIB is available on the OpenVMS VAX, OpenVMS AXP, ULTRIX, and
DEC OSF/1 AXP operating systems. The major applications are
DECwindows/Motif, DECwrite, the front-end input process, and
screen management (SMG).
IMPLEMENTATION OF THE HENKAN MODULE
The second layer is part of the Japanese input manipulator and is
called the henkan module or HM. (Henkan means conversion in
Japanese.) It does not handle I/O operation but accepts key input
from the application and converts it to a string in preediting.
Figure 8 summarizes the function of HM. An application passes the
key input to HM stroke by stroke. HM performs all Japanese
preediting operations; the application has no direct manipulation
of the input. Then the application gets the preediting string
from HM. Because HM does no I/O, it is independent of any
specific device. As a result, all applications, including
windowing systems, can use HM.
Figure 8 Henkan Module Function
I/O
KEY
====>
INPUT
+--+
+--------+
+--+ I/O
| | ==> | HENKAN | ==> | | ====> PREEDITING
| |
| MODULE |
| |
STRING
| |
+--------+
| |
| +--------------------+ |
|
APPLICATION
|
+--------------------------+
In addition, HM can access the resources of IMLIB. This feature
helps the unification of the Japanese input user interface and
reduces the number of similar product conversions. HM has another
significant capability. We defined the common (minimum)
application programming interface to potentially accept all
Japanese conversion engines and implemented "PLUGGS" in HM.
Therefore HM can use one or more engines for kana-to-kanji
conversion.
HM Mechanism Overview
HM is a tool that any application can use. An application passes
key input to HM by a normal procedure call. After HM processes
it, HM calls application routines with the processed result.
Because HM handles large string buffers, it dynamically
allocates/deallocates memory. To ensure that memory is retained,
we used a callback technique. (These techniques are described
later in the Callback Routines section.)
HM operates by key input as follows:
1.
HM gets a keycode from an application with procedure
arguments.
2.
HM gets the actions assigned to the key from IMLIB.
3.
If the key is not assigned to the Japanese input
operation, HM tells the application to process it
separately.
4.
If the key is assigned to the Japanese input operation,
HM processes it according to the actions.
5.
HM modifies the information to be displayed according to
the action and calls a registered callback routine to
update the screen.
HM passes the information that should be displayed on the screen
in an argument of the callback routines. The callback routines
are prepared by the application and registered into HM context at
the initialization of HM. This callback method makes the
application interface and data flow more easily.
Components
Figure 9 shows the composition of HM. The application interfaces
include both the C and the VMS binding interfaces for the OpenVMS
operating system.
Figure 9 HM Component Structure
+----------------------------------------+
|
SEVERAL APPLICATION INTERFACES
|
+----------------------------------------+
|
JAPANESE INPUT MANIPULATOR
|
+-------+----------------+---------------+
| IMLIB | ROMAJI-TO-KANA | KANA-TO-KANJI |
|
| CONVERTER
| CONVERTER
|
+-------+----------------+---------------+
The Japanese input manipulator performs all Japanese input
operation by using IMLIB, the romaji-to-kana converter, and the
kana-to-kanji converter. After it processes the input key, it
calls back the application routines. There are several types of
romaji-to-kana converters. We implemented a submodule
romaji-to-kana converter driven by a conversion table; a user can
change this table to another.
The kana-to-kanji converter module is a generalized Japanese
conversion library. Many Japanese conversion engines exist, and
each one is used differently. The kana-to-kanji converter loads
the interface routine that absorbs these differences dynamically
at the initialization of the HM context. It then processes the
conversion request with any engine.
Services
HM provides 17 library entries. In this section, we describe
three basic routines: HMInitialize, HMConvert, and
HMEndConversion.
o
HMInitialize. This routine creates a context for HM. It
accepts three callback entries, a user-defined data
pointer that would be passed to the callbacks, and an
item list for initial information as its arguments.
o
HMConvert. This routine sends a key to HM. The key is
represented as a 32-bit data (longword) that is generated
by a function HMEncodeKey from an escape sequence that
the keyboard sends or by a function HMKeysymToKeycode
from a keysym of the X Window System. IMLIB interprets
the keycode, and HMConvert performs a conversion in
accordance with the information. (A summary of what is
executed was given in the Mechanism Overview section.)
o
HMEndConversion. This routine aborts the conversion and
resets an internal status. It is used when the
application has to stop the input for a particular
reason, for example, if an application issues the cancel
request.
Callback Routines
HM requires three callback routines: start_conversion,
format_output, and end_conversion. They are used as follows.
o
start_conversion. This routine is called when the
conversion string input is started. The application
memorizes where the cursor is positioned.
o
format_output. This routine is called whenever the
information to be displayed has been changed. The
application updates the screen.
o
end_conversion. This routine is called when the input
string is determined. As a result, the application takes
the string passed in the argument of the last call of
format_output into its input buffer.
The user-defined data pointer, one of the arguments for
HMInitialize, is always passed to these callbacks. Since HM is
not concerned with its contents, the user can put any kind of
information into it.
HM is available on the OpenVMS VAX, OpenVMS AXP, ULTRIX, and DEC
OSF/1 AXP operating systems. This portability is due to the
module's independence from physical I/O. The major client
applications working on these operating systems are
DECwindows/Motif, Japanese SMG, and the front-end input process.
IMPLEMENTATION OF THE FRONT-END INPUT PROCESSOR
The front-end input process (FIP) for a dumb terminal supports
full operations for the Japanese string manipulation. FIP
is implemented on the following operating systems:
OpenVMS/Japanese/VAX version 5.5-2 or later versions and
OpenVMS/Japanese/AXP version 1.0 or later versions.
Full Operation Support
The original product can use FIP if the product's mechanism,
particularly its I/O operation and preediting function, does not
conflict with the FIP implementation. Some applications
conflict with the design of FIP due to the limitations of FIP and
its environment. For example, FIP does not detect the read
request that includes the NOECHO item code, so the application
that issues such a read request to the terminal driver (TTdriver)
cannot use FIP as a Japanese front-end input process. Also FIP
does not step into a process for the termination of a read
request simply because a read buffer that is defined by an
application has overflowed. FIP continues to communicate with
the TTdriver and a conversion engine to get the Japanese string
unless the terminate key is explicitly input. To overcome these
conflicts, we implemented a pseudo-driver named FIdriver to
intercept I/O requests from the application before they are
processed by the TTdriver.
FIP Mechanism Overview
FIP processes all Japanese input operations using HM. We supplied
the Digital Command Language (DCL) command, INPUT START/STOP for
activating/deactivating FIP. Once a user activates FIP from DCL,
it is available until the user logs out or the system is
deactivated.
Figure 10 shows FIP and its environment for the manipulation of
Japanese input. An application issues I/O requests to the
TTdriver to get user inputs, but FIP fetches the requests from
the TTdriver through the FIdriver. Then FIP starts to communicate
with the drivers and the Japanese string conversion engine to
pass the resultant string as well as preedits to a screen.
Figure 10 FIP Environment for Manipulation of Japanese Input
+-------------+
| DICTIONARY |
+-------------+
^
|
|
v
+-------------+
| CONVERSION |
|
ENGINE
|
+-------------+
^
|
THE INPUT |
| THE RESULT
STRING
|
| STRING
|
v
+---------------+ SEND START REQUEST +----+ +-------------+
| AN
|--------------------> |MAIL|--|
|
| APPLICATION
|
|BOX | |
FIP
|
|
|
+----+ |
|
|
|
SEND START CONFIRM |
|
|
| +----+
/---------------|
|
+---------------+--|MAIL|<----/
/-->+-------------+
|
^
|BOX |
/
^
^
|
|
+----+
/
|
|
$QIO READ
|
|
WRITE BACK THE RESULT /
|
|
REQUEST
|
+----------------------------------+ |$QIO | AST
|
/
| |
|
|
+-------------------------/
| |
|
v
v
$QIO READ/WRITE REQUEST
| v
|
+-------------------+
+--------------+
|
|
|
|
|
| TTDRIVER
| HOOK |----------------------->|
FIDRIVER
|
|
|
| INTERCEPT OF READ $QIO |
|
+-------------------+
+--------------+
|
^
v
|
+----------------+
| TERMINAL /
|
| INPUT DEVICE |
+----------------+
The sequence of the front-end input process follows.
1.
An application creates a front-end input process.
2.
A front-end input process exchanges packets with an
application through its mailbox.
3.
An application issues a queued I/O ($QIO) read request to
the TTdriver.
4.
The FIdriver intercepts the request and passes the
information to FIP as a packet.
5.
FIP issues a $QIO read request to the TTdriver to get
input strings for conversion.
6.
A user inputs a key from a terminal. FIP receives the
input and decides whether or not to call a routine of the
conversion engine. If an input key is recognized as one
of the conversion keys, FIP calls the routine and passes
the input strings. If not, FIP issues a $QIO write
request to the TTdriver to echo an input character.
7.
A conversion engine receives a string and converts it to
the Japanese string.
8.
A conversion engine returns the result to FIP.
9.
FIP issues a $QIO write request to the TTdriver to
display the resultant string from the engine and arranges
the current editing line.
10.
Steps 5 to 9 are repeated.
11.
Once a user inputs the Terminate key of an application's
request, FIP recognizes it as a terminator and returns
the entire resultant string to the FIdriver as a write
packet.
12.
The FIdriver sends the result string and I/O status to an
application.
13.
An application accepts the converted string. After
executing its internal process, it issues another $QIO
read request to the TTdriver. (Return to step 3.)
FIdriver.
The FIdriver is a pseudo-driver that intercepts $QIO
read requests from an application to the TTdriver. Functioning as
a bridge between terminal read requests and FIP, the FIdriver
gets a read request, passes its information to FIP, and maintains
it. When FIP returns the completion message with its processed
Japanese string, the FIdriver validates it and completes a user's
read request as if the TTdriver had returned it. Thus the
user/application can get the Japanese string without modification
for Japanese input method.
The FIdriver has other notification functions for exception
handling such as logout, cancel, or abort.
Front-end Input Process Operations. All the operations in the
front-end input process are driven by the mailbox event, the
FIdriver event, and the key event. Figure 11 shows the functional
structure of FIP.
Figure 11 FIP Functional Structure
+----------------------------------------+
|
FIP EVENT MANAGEMENT
|
+-------------------+--------+-----------+
|
KEY EVENT
|
|
|
+-------------------+ FIDRIVER | MAILBOX |
| ACTION CONTROL
| EVENT
| EVENT
|
+----------+--------+
|
|
| TERMINAL | HENKAN |
|
|
| ACTION
| MODULE |
|
|
+----------+--------+----------+---------+
The following operations in the front-end input process
correspond to these three events.
o
Mailbox Event. The mailbox event provides communication
with an application. FIP issues a read request to its own
mailbox. The mailbox event notifies FIP of the arrival of
a message from an application. When an application sends
a start request to the FIP mailbox, the mailbox event is
set so FIP starts to initialize its environment. Also FIP
terminates itself at the time a stop request message is
delivered to its mailbox.
o
FIdriver Event. The FIdriver event provides
communication with the FIdriver. The FIdriver intercepts
a request from an application to the TTdriver and creates
a packet for FIP. FIP issues a read request to the
FIdriver, and this event is set when a packet is
delivered. A request is categorized in three types: read
request, cancel request, and disconnect request.
o
Key Event. The key event provides communication with the
TTdriver. FIP issues a $QIO read request to the TTdriver
byte by byte. All the input from a keyboard is recognized
as a key event in FIP. Once a key event is set in FIP,
FIP examines the key sequence in a read buffer.
If the input is in the range of a terminator mask, FIP terminates
a read operation from the TTdriver and writes back the resultant
string and I/O status block to the FIdriver as a write packet. (A
terminator mask is defined in the $QIO read request from an
application.)
If the input key is a conversion key, FIP calls a conversion
engine and gets the resultant converted string. Then FIP issues a
write request to the TTdriver to display the updated string.
If the input key is a printable character, FIP updates the
contents of its internal buffers defined in the context and
issues a write request to the TTdriver to echo the character.
If the input key is for line editing, for example, to delete a
line or a word or to refresh a line, FIP emulates the
line-editing function of the TTdriver so its editing function is
executed.
FIP stores all user input and read-request information from an
application in its internal buffers and database. The buffers
contain the codes of user input and corresponding video
attributes to display. The database contains item codes in a read
request, channel numbers to connect other devices, and so on.
FIP creates a new database when the updated read request from an
application is delivered, in other words, when the FIdriver event
is set. Also, FIP adds the ASCII code and an attribute of the
updated user input into buffers when a user inputs, that is, when
the key event is set.
CLIENT/SERVER CONVERSION
The use of a client/server conversion has two advantages: (1) It
reduces the required resources for language conversion by
distributing some components to other systems, and (2) It
presents an environment that shares a common dictionary.
All procedures for the Japanese conversion require large system
resources such as CPU power. A user can place the conversion
information server (CIserver) and a dictionary on a remote node
and call some functions of the CIserver client library to get the
resultant string. In this way, a local system saves its resources
while the remote server processes the conversions.
In addition, many users can access a common dictionary on the
specific remote node. It is possible for any local user to access
a dictionary on a remote node if the CIserver on the node is
active.
CIserver
The object name is "IM$CISERVER". The CIserver initializes itself
by finding the name of a transport protocol in a logical table.
It then creates corresponding shareable images, maps its required
routines, and waits for a connect request from a client. The
CIserver communicates with its client via a mailbox at the
transport level. The server sets the asynchronous system trap to
the mailbox and reads a message in it such as a connect request,
a disconnect request, a connect abortion, or a client's image
termination. The CIserver can identify the connection to a client
and specify a conversion stream in the connection.
CIserver Client Library. The client library presents programming
interfaces. These are callable routines that execute various
string manipulations and operations for the Japanese conversion.
The CIserver client library is located between an application and
the CIserver body.
Input Method Control Program (IMCP). IMCP is a command line
interface to customize the CIserver environment. A user sets
proxy to a Japanese system dictionary at a remote node on the
network, and IMCP administrates a proxy database. A user can
confirm the status of the server at a command line and can shut
down the server from the IMCP interface.
Other Servers. HM has a conversion engine dispatcher that can
dynamically select from several Japanese conversion engines. HM
now serves the CIS (CIserver, Digital Japan), the Wnn (Omron
Company), the Canna (NEC), and the JSY (Digital Japan) engines.
Therefore, an application that uses HM as the Japanese conversion
interface can select its preferred engine.
EXTENSION IN THE FUTURE
In this section, we describe the possibilities for
internationalization of FIP, HM, IMLIB, the CIserver, and the
FIdriver. Although our approach does not provide a multilingual
input method, it does provide an architecture that can be used
for any language.
FIP has a multibyte I/O operation that can be applied to other
two-byte languages. In addition, all the read/write
communications among FIP, the FIdriver, and the TTdriver proved
able to handle one-byte languages such as English. Also, IMLIB
can expand its keybinding system for conversion of other
languages, and HM can add the interfaces for conversion engines
of other languages if such engines are prepared.
SUMMARY
The Japanese input method is a complex procedure involving
preediting operations. Various keybinding systems and
manipulators accelerate input operations. Our approach for the
Japanese input method allows an application three choices: (1) An
application can use a front-end input processor to perform all
operations including I/O. (2) An application can use the
henkan module and implement I/O operation itself. (3) An
application can execute all operations except keybinding, which
is handled by an input method library.
ACKNOWLEDGMENTS
We want to express our appreciation to Katsushi Takeuchi of the
XTPU development team for his initial designing and prototyping
of IMLIB and some implementation of FIP, and Junji Morimitsu on
the same team for his initial implementation of IMLIB and its
compiler. Also, we wish to thank Makoto Inada on the DECwindows
team for his implementation of HM; Hitoshi Izumida, Tsutomu
Saito, and Jun Yoshida from the JVMS driver team for their
contribution toward creating the FIdriver; and Naoki Okudera for
his implementation to the entire CIserver environment. As a final
remark, we acknowledge Eiichi Aoki, an engineering manager of ISE
Japan, and Hirotaka Yoshioka in the ISA group for their
encouragement in writing this paper.
REFERENCES
1.
Guides to the X Window System Programmer's Supplement for
Release 5 (Sebastopol, CA: O'Reilly & Associates, Inc.,
1991).
2.
Standard X, Version 11, Release 5 (Cambridge, MA: MIT X
Consortium, 1988).
3.
K. Yoshimura, T. Hitaka, and S. Yoshida, "Morphological
Analysis of Non-marked-off Japanese Sentences by the Least
BUNSETSU's Number Method," Transactions of Information
Processing Society of Japan, vol. 24 (1983).
4.
K. Shirai, Y. Hayashi, Y. Hirata, and J. Kubota, "Database
Formulation and Learning Procedure for Kakari-Uke Dependency
Analysis," Transactions of Information Processing
Society of Japan, vol. 26 (1985).
5.
IMLIB/OpenVMS Library Reference Manual (in Japanese) (Tokyo:
Digital Equipment Corporation Japan, Order No. AA-PU8TA-TE,
1993).
6.
User's Manual for Defining User Keys in IMLIB (in Japanese)
(Tokyo: Digital Equipment Corporation Japan, Order No.
AA-PU8UA-TE, 1993).
TRADEMARKS
The following are trademarks of Digital Equipment Corporation:
DEC OSF/1 AXP, DECwindows, DECwrite, Digital, OpenVMS AXP,
OpenVMS VAX, ULTRIX, and VMS.
Motif and is a registered trademark of the Open Software
Foundation, Inc.
Unicode is a trademark of Unicode Inc.
UNIX is a registered trademark of UNIX Systems Laboratories, Inc.
X Window System is a trademark of the Massachusetts Institute of
Technology.
BIOGRAPHIES
Hiroyoshi Baba Hiroyoshi Baba is an engineer in the Japanese
Input Method Group in Digital Japan, Research and Development
Center. He is currently developing the Japanese front-end input
system on OpenVMS VAX and OpenVMS AXP and the Japanese language
conversion server system. He received a B.S. (1989) and an M.S.
(1991) in electronics engineering from Muroran Institute of
Technology, Japan. He joined Digital in April 1991.
Takahide Honma A senior software engineer, Takahide Honma leads
the Japanese Input Method Group. He joined Digital in 1985 as a
software service engineer. He has worked on systems such as
real-time drivers, network system (P.S.I.), and database on VMS
and was a consultant to customers. At the same time, he also took
the role of a sales advisory support engineer. Since 1990, he has
been with Research and Development in Japan and has worked on the
Japanese input method. He has an M.S. (1983) in high-energy
physics from Kyoto University and is a member of the Physics
Society of Japan.
Kuniaki Takizawa Kuniaki Takizawa is an engineer with Digital
Japan, Research and Development Center and is a member of the
Japanese Input Method Group. He joined Digital in April 1991 and
is currently developing and porting the henkan module and the
input method library (IMLIB) on OpenVMS, ULTRIX, and OSF/1. He
graduated from the University of Electronic Communications
(Denki-Tsushin University) in Japan in 1991. His speciality area
was the structure of operating systems.
=============================================================================
Copyright 1993 Digital Equipment Corporation. Forwarding and copying of this
article is permitted for personal and educational purposes without fee
provided that Digital Equipment Corporation's copyright is retained with the
article and that the content is not modified. This article is not to be
distributed for commercial advantage. Abstracting with credit of Digital
Equipment Corporation's authorship is permitted. All rights reserved.
=============================================================================